This paper presents the results of a study on interplot competition in cereal variety trials, which consisted in analysing official trials and in carrying out specific experiments. The results show that small and wholly harvested plots undergo marked competition effects related to plant height. These effects are largely, but not completely, removed by using two guard rows at both sides of plots. The use of height differences as a covariate in the analysis is an other efficient way of controlling competition. The results suggest that neighbour-balanced designs are not sufficient to remove bias but that they allow competition to be taken into account in the analysis more efficiently.